DeepSeek-R1 is not an architecture paper: it is a post-training report on a 671B/37B-active MoE model whose base is DeepSeek-V3, already in the atlas. Everything the report discloses about the block fits in Table 4 — 'Architecture: MoE', 671B total, 37B activated — so every deeper spec (attention, experts, context, vocabulary) is deliberately unknown here rather than imported from the V3 record. The substance is the two-stage reasoning recipe: R1-Zero showed pure GRPO with rule-based rewards, no SFT at all, elicits long chain-of-thought and the emergent 'aha moment'; R1 adds a cold-start CoT SFT before RL, then a rejection-sampled SFT round and a second RL stage for helpfulness/harmlessness. The report also established the pattern of distilling reasoning into small dense models, which several later open-weight families in the gallery follow.
What the report claims is novel
R1-Zero: first demonstration of pure RL (no SFT) eliciting strong reasoning, including the emergent 'aha moment' [Abstract / §2.1 / §3.1]
GRPO-based RLVR with rule-based rewards avoiding reward-model bottlenecks [§2.2]
Distilling reasoning capabilities into small dense models (1.5B–70B) with strong results [§2.4 / §3.4]
Open-sourcing R1-Zero, R1, and six distilled dense models [Abstract]
Atlas assessment: Architecturally this record adds nothing new to the atlas — the report states the base is DeepSeek-V3 (already a record) and discloses no block details. Its value is as the reasoning-specialised exemplar of the sparse-MoE line: the first major open report showing rule-based RLVR alone (without SFT) can elicit chain-of-thought reasoning, plus the distillation-of-reasoning pattern that several later atlas entries adopt.
Relation to the atlas
Lineage in
MoE base architecture (671B total / 37B activated) from DeepSeek-V3
GRPO reinforcement learning from DeepSeekMath (Shao et al.)
Influence out
deepseek-v3-2: Group Relative Policy Optimization (GRPO) as the RL training algorithm